At their core, robotic floor cleaners integrate at least two primary functional systems: (1) a cleaning mechanism, which cleans the floor in the area where it is placed or moved across, and (2) a mobile robotic platform, which autonomously moves the cleaning mechanism across the floor to different places. Both of these functional systems must work adequately for the robot to be effective at cleaning.
While both functional systems set requirements and constraints on the design of the overall robot, the challenge of developing a mobile robotic platform that can autonomously move around in nearly an infinite variety of highly unstructured environments (e.g., people's homes) tends to be the dominant consideration and has had a significant effect on the design of home cleaning robots to this date.
In terms of robotic floor cleaners currently available, the constraint is so great that the vast majority of units manufactured to this date follow the same general form factor. The mobile robot platform is contained within a shell having a circular base, similar in shape to a hockey puck but much larger than that. Two independently controlled drive wheels are set within the circle on opposite sides of the robot. The wheels are located along the center axis of the circle bisecting the forward and rear halves of the robot.
In addition, the mobile robot platform has one or more caster wheels for support at the forward and/or rear ends of the robot to provide lateral stability and act as part of the robot's suspension. Some designs use only one caster in the front, but distribute the weight to be heavier in the front to keep the robot from tipping backward.
The circular shape makes the robot much easier to navigate around obstacles and along walls. With the wheels fully nested within the circle and placed along the center axis, the robot can effectively turn in place to change its heading without the sides of the robot hitting any exterior obstacles.
The cleaning robot design also allows contact sensors (e.g., located on a bumper) and proximity sensors (IR sensors) to be placed along the outer sides of the robot to detect obstacles and follow along walls and furniture. In some designs, the bumper may extend outside the boundary of the circular base as a means for feeling for walls and obstacles as the robot turns. The current cleaning robot design also includes drop sensors beneath the robot for detecting drop offs in the floor before the robot drives over a hazard.
Examples following this design framework include robot vacuum cleaners such as Eletrolux®'s Trilobite™, iRobot®'s Roomba®, Yujin's iClebo, Samsung®'s Hauzen® robots, as well as floor scrubbing robots such as iRobot®'s Scooba™ robot. The down-side of the presently available solution for the mobile robotic platform is that the current configurations limit the size, reach and effectiveness of the cleaning mechanism.
In the typical robotic floor cleaner approach, the primary cleaning mechanism (such as a vacuum or beater brush) is designed to fit entirely within the footprint of mobile robot platform. Given that the dominant platform shape is circular for most robotic floor cleaners, the cleaning apparatus necessarily has be narrower than the robot itself, cover a smaller area of the floor relative to the size of the overall robot, and is not able to directly reach all the way to walls and into corners. This is particularly sub-optimal for cleaning mechanisms such as vacuums, brushes and other devices, which tend to be rectangular in shape and don't match well with the geometry of a circle.
To compensate for this limitation, designers of the robotic cleaners have added an “edge cleaning” feature in the form of a small side spinning brush that reaches out from the side of the robot. The small side spinning brush attempts to draw debris into the path of the cleaning apparatus, although this has limited effectiveness and often needs to be replaced due to wear.
The typical placement of the wheels and casters to support a circular robot platform places additional constraints on the cleaning mechanism. As one limitation, the cleaning mechanism cannot extend to an area where there is a wheel or caster, further limiting its size and configuration. Additionally, the robot usually requires a more complex suspension system to keep the cleaning mechanism in contact with the floor to be effective in cleaning, while at the same time maintaining primary contact between the floor and the wheels and casters in order for the robot to be effective in driving over the floor surface and over small obstacles.
A number of variations on the typical robot floor cleaner configuration have been proposed in an attempt to reduce the constraints on the cleaning apparatus and increase its effectiveness, but these solutions still prioritize the mobility and function of the robotic platform over the function of the cleaning mechanism itself.
Products with a non-circular shape, such as The Shaper Image®'s oval shaped eVac® robotic vacuum, have been introduced to the market. In this oval-shaped robotic vacuum, the front leading edge was flattened to allow the vacuum to reach parallel to the wall in front of the robot, but the cleaning mechanism was still nested within the shell of the cleaning robot and did not extend to the sides.
Other robotic cleaner designs have been disclosed which combine a mobile robot platform with a cleaning apparatus that is partially or fully extending past the footprint of the mobile robot platform. These examples include a robot with a flexible tail extending outside the shell of the cleaning robot, Proctor and Gamble (“P&G”), robots with a trailing clearing module (SC Johnson and Minolta), as well as a cleaning module that is movable relative to the mobile robot platform (Minolta).
U.S. Pat. No. 6,779,217, assigned to P & G, discloses a mobile robot platform which utilizes the standard circular design, but also includes a flexible “appendage” in the form of a triangular tail, where the triangular tail holds a cleaning cloth. The advantage of this design is that it can reach into corners with the extended tail, as well as clean along the edges of furniture and walls. However, the extended reach of its cleaning abilities works only when the cleaning robot turns in place to “sweep” along the radius of the turn. While the “appendage” approach provides a beneficial supplementary function for catching extra dirt and dust, this approach does not overcome the limitation of the primarily cleaning mechanism being placed within the footprint of the mobile robot platform. For example, if the cleaning robot drives along the side of a wall, it will still not clean a majority of the gap between the primary cleaning mechanism and the wall, and will only do so in the limited area where the cleaning robot makes a turn.
SC Johnson, in U.S. application Ser. No. 10/218,843, disclosed a configuration which combines a circular mobile robot platform with a trailing external cleaning pad that could hold a cleaning cloth or other material. This cleaning robot design allows the external cleaning pad to be as wide as the cleaning robot, and provides that the edges of the cloth can extend past the width of the pad to reach along walls and into corners. The drive system and sensors for the robotic cleaner would be in the front circular section as part of the mobile robot platform. The limitation of this design is the larger size and longer shape of the combined form factor, which limits how well the robot would be able to navigate in tighter spaces. This robotic cleaner configuration would have the advantage of the standard circular design for turning along walls and, to some extent, maintain the benefits of being able to bump and turn to get around forward obstacles. However, the extended length of the cleaning robot would provide challenges for turning in tighter spaces, as well as for getting in and out of cluttered environments, such as between chair and table legs which are clustered together. The extra length of the trailing pad would prevent the cleaning robot from navigating into spaces a robot with just the circular section could enter.
In U.S. Pat. No. 5,894,621, Minolta disclosed a similar robotic cleaner configuration of a cleaning pad trailing the mobile robot platform, where the pad would be larger than the robot body to allow greater access to walls and side cleaning. However, this configuration would still have the same limitations of navigating in tight spaces given the overall length and distance from the wheels to the cleaning pad.
In U.S. Pat. No. 5,720,077, Minolta disclosed another cleaning robot design where the cleaning mechanism is external to the mobile robot platform. The cleaning robot expands its reach by making both (1) the cleaning mechanism's position adjustable to the mobile cleaning robot and (2) making the mobile robot platform drive wheels change the primary axis of direction, relative to the cleaning mechanism, in order to drive it in different orientations. This robotic cleaner design offers a greater degree of flexibility for cleaning in different spaces, but comes at a price of significantly more cost and complexity. Specifically, this robotic cleaner design includes more moving parts and more sensors to judge situational conditions and control the position of the cleaning mechanism, so this will likely result in a physically larger robot. While this may be appropriate for a commercial robotic cleaner for large office and commercial spaces, this disclosed cleaning robot design would not fit the requirements of a consumer robotic cleaner needing to clean in tight spaces, such as around a kitchen table that is positioned close to one or more walls, which also includes a number of chairs, as well as in places deep under low furniture.
The cleaning pattern and navigation strategy of consumer robots is also an area in need of improvement. The vast majority of current consumer cleaning robots on the market utilize a random or semi-random navigation scheme for controlling the robot's driving behavior. This is primarily because in the past there has not been an effective and low-cost navigation solution that can track the robot's position and guide it to intelligently cover the desired area for cleaning the floor.
Instead, cleaning robots normally rely on a relatively simple set of behaviors and algorithms that combine driving, obstacle detection and avoidance, wall following and random variables in an attempt to “bounce” the robot around the floor space. The rationale is that over enough time, the robot will reach all locations in the cleanable area just through the process of randomly exploring the room.
This approach has several major limitations. First, the robot must clean for very long periods of time to reach full coverage of a room or other designated area of a home. As noted in U.S. Pat. No. 6,076,025, the rate of new area covered drops significantly the longer the robot operates. Because the robot has little to no prior knowledge of where it has cleaned, it continues to clean over areas that it has already cleaned before as opposed to focusing on areas to it has not yet reached. As more area of the room or area of the home is covered, the random method works against the robot, as the robot spends proportionally more and more time cleaning areas it has already cleaned. For a typical large room or area in a home, the robot runs out of battery power before it's able to reach all locations in the target area.
In U.S. Pat. No. 6,076,025 (“the '025 patent”) assigned to Honda, Honda discloses an enhanced approach by having the robot periodically clean in an outward spiral pattern during the course of randomly navigating through a room. This approach has the benefit of filling in more area in different locations and improving the efficiency curve relative to a pure random approach, but the same core dynamic exists in terms of making the robot becoming increasing inefficient as more of the room is covered.
As another limitation, cleaning strategies that are described as random or semi-random do not normally result in a random distribution of cleaning coverage across a typical room or area of a home. In a completely empty room where there are only walls and not interior obstacles, each area of the floor will have an equal chance of being cleaned within a certain amount of time. So on any given run, one part of the room is just as likely as another part of the room to get cleaned. As more runs are made or repeated, the odds increase that a specific part of the room will be cleaned.
However, rooms and areas in homes are not empty, and instead have a variety of interior obstacles which the robot must navigate around. Additionally, these obstacles are not randomly dispersed. Rather, large obstacles, namely furniture, are clustered together and form uneven barriers of entry for a robot to navigate through. A common example includes a dining table and chairs, which create a “forest” of furniture legs concentrated in a certain area of a room. Another common example is a living room furniture set, such as a long couch, coffee table, side chairs and side tables clustered in a pattern, such as U-shape or L-shape configuration.
In theory, the random approach gives infinite opportunities for the robot to find openings in the room between obstacles by hitting them in all points along their perimeter at virtually all angles. However, this process takes time for enough permutations to take place and for a large percent of those permutations, the robot may be frequently blocked and/or deflected away from areas occupied by these obstacles. The net result is that the robot following a random approach is “corralled” away from dense or blocked areas, and tends to stay into more open spaces. This often causes the robot to significantly over clean some areas of the room or areas of the home while under cleaning other areas. In other words, a random and/or semi-random guided robot can systematically avoid and under-clean certain areas, over and over. This results in poor cleaning of those areas, inefficient use of energy, as well as excess wear on the robot and potentially of the floor covering in the over-cleaned areas.
These limitations of a random or semi-random approach, the general low efficiency in reaching all areas of the floor, and the systematic pattern of repeatedly missing certain parts of the floor area, are counter-productive to the primary cleaning function of the robot. This is especially true for robotic cleaners which use a consumable material for cleaning that has a limited period of use. As an example, for robotic cleaners such as described in this invention which uses a wet cleaning pad, the dispersion of fluid on the floor would be concentrated in certain areas of the room, possibly to the point of pooling and leaving streaks and residue spots on the floor. At the same time, the pad dries up as it cleans, meaning areas that are not reached until later will not be wet or cleaned at all.
To address these limitations, companies have proposed and in a few instances developed robots which use a systematic or semi-systematic cleaning strategy. These robots generally follow some pre-defined cleaning behavior or pattern, such as crossing the floor in parallel rows, to provide a more even and controlled method of cleaning the floor area. For example, Samsung U.S. Pat. No. 7,480,958 discloses these types of robots.
For open areas of a room, a parallel row pattern can be much faster at covering a large amount of area in much less time than a random approach. These patterns can also have the benefit of using the driving pattern to probe for open areas amidst obstacles that lie in the path of the rows. In the case of a robot cleaning in a pattern of parallel rows, the ends of the rows provide an opportunity to probe for open spaces between obstacles in a fast and systematic way. For example, if the direction of the rows is roughly perpendicular to a boundary formed by a cluster of furniture, then the robot has the opportunity to attempt to pass into openings in the cluster on each returning pass. By controlling the spacing on the rows and using heuristics for when to attempt to follow around an obstacle, the robot can be much more methodical about finding spaces in between obstacles that a random approach would initially miss.
The challenge with systematic cleaning patterns is that the structure of the pattern may be too rigid to adapt enough for successfully covering all areas in a complex environment. While a random cleaning strategy over time will allow the robot to find its way almost everywhere just by chance, although not efficiently or with equal success, a systematic strategy alone cannot guarantee full coverage. The problem is the pattern generally has one preferred direction and follows rules about which way to proceed when it has multiple directions to select from. So in the example of the parallel rows, the robot may end up cleaning on side of a kitchen aisle, but miss the other side.
To make structured patterns work in unstructured environments, more capabilities have to be added to the system to track the robot's position, keep a map of where it has cleaned and identify areas that have not been reached. To date, very few systems have been developed that would be suitable for a home cleaning robot. As one example, the Samsung® Hauzen® vacuum cleaning robot uses a camera to create a map of where it cleans to provide some of these functions, but this product requires complex image recognition software and more expensive sensors and internal computing hardware built into the robot. To date, this product is significantly more expensive than even the premium version of the leading models for robotic floor cleaners, and accounts for a very small share of the market. Evolution Robotics® has alternative vSLAM® localization system utilizing a camera and visual pattern recognition to construct a map of the environment, recognize where the robot is, and guide the robot, but this system too requires generally more processing and memory than is on the current generation of cleaning robots. Evolution's patents disclosing vSLAM localization system information include U.S. Pat. Nos. 7,272,467; 7,177,737; 7,162,338; 7,145,478; 7,135,992 and 7,015,831. For the vast majority of robots sold to consumers, random or semi randomly cleaning remains to be the standard.